Various electronic systems exist for measuring extremely small differences in sensor measurements, such as temperature, for use in biological and physical analysis. It is known in the art that active and passive electronic components in such systems are subject to time and temperature drift, and that under normal operating conditions, the amplitude of time and temperature component drift is typically much greater than the amplitude of other inaccuracies generated by system components, such as amplifier noise voltage, noise current, and resistor noise. Consequently, component time and temperature drift are significant limiting factors to high resolution measurements, such as temperature difference measurements. To address the problem of component drift in electronic measurement systems generally, various approaches to compensate for drift have been devised.
For example, U.S. Pat. No. 5,253,532 (Kamens); U.S. Pat. No. 5,042,307 (Kato); U.S. Pat. No. 4,611,163 (Madeley); and U.S. Pat. No. 3,831,042 (La Claire) disclose electronic measurement systems (principally directed to pressure sensing, in the preferred embodiments) which include additional hardware components that change their electrical resistance, or other electrical parameters, with ambient temperature, in such a way as to compensate for thermal drift in measurement systems to which they are electrically connected. While such hardware compensation systems provide some compensation for thermal drift inaccuracies, they do not compensate for component drift over time, particularly the drift of sensors, such as thermistors. This would be sufficient to preclude temperature difference measurements, with resolution on the order of micro-degrees centigrade, if these techniques were applied to that purpose. Additionally, such hardware based compensation techniques do not readily compensate for component drift, resulting from the combined time drift characteristics of multiple system components, located at different parts of the system, with different thermal drift characteristics, and subject to non-uniform aging. In any case, the ability of the above hardware based compensation systems and techniques to compensate for system thermal drift are limited by the extent to which the particular technique tracks with thermal drift of the overall system, over time and temperature. Consequently, such techniques would not provide sufficient compensation for component time and temperature drift to permit differential temperature measurements, with resolution on the order of micro-degrees centigrade, if these techniques were applied to that purpose.
Other hardware compensation techniques, such as disclosed in U.S. Pat. No. 5,616,846 (Kwasnik); U.S. Pat. No. 5,171,091 ((ruger et al.); and U.S. Pat. No. 5,132,609 (Nguyen), require a time and temperature stable reference signal, and U.S. Pat. No. 5,351,010 (Leopold et al.) requires the use of precision analog amplification hardware and costly time and temperature stable resistors. The required precision analog components in these systems results in increased cost, complexity, and power consumption. Moreover, these systems do not compensate for time drift of passive components, such as thermistors, which would be sufficient to preclude temperature difference measurements, with resolution on the order of micro-degrees centigrade, if these systems were applied to that purpose.
Additionally, U.S. Pat. No. 5,162,725 (Hodson et al.); U.S. Pat. No. 5,065,613 (Lehnert); U.S. Pat. No. 4,958,936 (Sakamoto et al.); and U.S. Pat. No. 4,464,725 (Briefer) describe electronic measurement systems which compensate for thermal drift, and other system inaccuracies, by utilizing a computer, and memory for storing known temperature behavior of a measurement system, at various calibration temperatures. That is, system inaccuracies due to temperature drift are recorded at specific calibration temperatures. This stored temperature behavior is then used to interpolate system inaccuracies due to thermal component drift at operational temperatures within the calibration range. This has been accomplished by using mathematical formulae to model thermal offset curves (e.g., using a parabolic interpolation, such as the LaGrange method, to plot offset curves, based upon discrete offset measurements, at discrete ambient temperatures), and then during normal operation, using a said formula, with a current ambient temperature measurement, to determine expected circuit offsets for the current ambient temperature, so that actual system measurements during normal operation can be adjusted for the effects of said expected circuit offsets. The prior art measurement systems, which utilize a computer, can provide time and temperature compensation based only upon the most recent reference calibration data, the acquisition of which requires that the system be cycled through an entire temperature range, and is sufficiently time consuming to prevent, or significantly interrupt, normal system operation. Other techniques that utilize software compensation, such as disclosed in U.S. Pat. No. 4,532,601 (Lenderking et al.), as well as in U.S. Pat. No. 4,464,725 (Briefer, referred to above), require the use of a time and temperature stable reference signal, which increases cost, complexity, and power consumption. U.S. Pat. No. 4,959,804 (Willing) utilizes time and temperature stable passive components, which, even with costly bulk metal foil, or wirewound, resistors, would not provide the accuracy necessary in temperature difference measurements, with resolution on the order of micro-degrees centigrade, if this technique were applied to that purpose. Such time and temperature stable bulk metal foil resistors (e.g., manufactured by Vishay Electronics Foil Resistors, of Malvern, Pa.) and wirewound resistors, such as manufactured by Dale Electronics, of Norfolk, Nebr., are one to two orders of magnitude more expensive than standard metal film resistors, which provide comparable time stability, but are not nearly as temperature stable. Furthermore, U.S. Pat. No. 4,959,804 (Willing, referred to above) updates a previously recorded temperature curve according to the two endpoints of the curve, thereby ignoring variations that might occur at intervening points along the curve, over time, as well as time drift in temperature measuring thermistors, which drift sufficiently over time to invalidate temperature difference measurements, with resolution on the order of micro-degrees centigrade, if it were applied to the purpose of high resolution differential temperature measurements. U.S. Pat. No. 4,651,292 (Jeenicke) relies on updating a point on a measurement curve, requiring, however, the restriction that the sensor curve characteristic be linear (not the case with temperature sensors, such as thermistors, and not sufficiently so, to provide resolution on the order of micro-degrees centigrade, even with known thermistor linearization techniques) and that ambient temperature measurements not drift with time, to the extent that measurement accuracy would be affected, making this technique unsuitable for a differential thermometer with resolution on the order of micro-degrees centigrade, if such a technique were to be applied to that purpose.
In the above approaches, as they would relate to a temperature difference measurement system, utilizing a pair of thermistors (e.g., in a thermistor-resistor bridge arrangement), it is relevant to note that differences in thermistor resistance-temperature curve characteristics, between two thermistors, result in a difference in the two thermistor resistances, throughout an ambient temperature range, that varies significantly with ambient temperature. For instance, YSI 460 series "Super-Stable Thermistors", manufactured by YSI, Incorporated, of Yellow Springs, Ohio, are characteristic of well matched, commercially available thermistors, and are matched to within 0.05.degree. C. of each other, between 0.degree. C. and 50.degree. C. This means that within the 0.degree. C. to 50.degree. C. range of operating temperatures, the difference in thermistor resistances may change by as much as 0.001.degree. C., relative to each other, for each ambient temperature change of 1.degree. C., a significant amount in measurements intended to resolve temperature differences on the order of micro-degrees centigrade. The above approaches, as they would relate to a temperature difference measurement system, utilizing a pair of thermistors, do not provide a means to compensate for this effect. Additionally, in order to minimize common mode amplifier error, the use of bipolar power to the measurement bridge is often preferred in the above prior art, as are high precision amplifiers, which typically require bipolar power, resulting in added cost and complexity, compared to a single-ended power supply architecture.
In U.S. Pat. No. 5,295,746 (Friauf et al.), directed specifically to the technical field of temperature difference measurement, with resolution on the order of micro-degrees centigrade, it is pointed out that a number of digital thermometers exist, which, however, have accuracy limitations on the order of one hundred milli-degrees centigrade. U.S. Pat. No. 5,295,746 addresses some of the limitations of the prior art, in this respect, by using a computer to maintain a thermistor-resistor bridge in a balanced state, to provide a means for adjusting thermistor power dissipation and to null out thermally generated offsets in the system's analog to digital converter, digital to analog converters, and amplifiers, for a given thermistor power dissipation, for the current ambient temperature. However, this requires that the bridge circuit be balanced with extreme accuracy, requiring the addition of two digital to analog converters to the circuit, preferably employing high resolution, in order to achieve high resolution temperature difference measurements, resulting in added component count, cost, and power consumption. Additionally, no means is provided to compensate for time drift of thermistors, which typically amounts to ten or more milli-degrees/year (e.g., YSI 44018, manufactured by YSI Incorporated, of ellow Springs, Ohio), a significant figure in temperature difference measurements, intended to approach micro-degree centigrade resolution. Additionally, software calibration of this system, for a current ambient temperature, is undertaken at the time when temperature difference measurements are undertaken, yet requires that bridge thermistors be completely powered down first, so that there is zero voltage potential across the bridge during calibration. That is, system calibration followed by continued operation in the temperature difference measurement mode must be undertaken by first powering down the bridge thermistors, and then powering them up again. Due to self-heating properties of thermistors, when a voltage is placed across thermistors, time is required for the thermistors to reach equilibrium with ambient temperature, which they do asymptotically. In cases where temperature difference measurements on the order of micro-degrees centigrade are to be resolved, this powering down and then powering up of the bridge, until the thermistors are within micro-degrees centigrade of equilibrium, adds significant time to the calibration process. Since each calibration offset measurement is associated with a specific ambient temperature, a calibration measurement (and, therefore, an ambient temperature) must be associated with each temperature difference measurement. Calibration measurements performed before and after a measurement run can conceivably be used to interpolate linear changes in ambient temperature with time, during a measurement run, but this places an unrealistic limitation on a measurement system which desirably operates under normal atmospheric conditions, in which ambient temperature changes may not be linear with time. Therefore, a calibration measurement must be undertaken for each temperature difference measurement, in which ambient temperature may not have undergone a linear change, thus adding significant time to the measurement process. Similarly, U.S. Pat. No. 5,351,010 (Leopold et al., also mentioned above) requires that current be reversed through zero, in resistive sensors, for each calibration, as well as requiring precision circuitry, that increases cost, complexity, and power consumption.
Additionally, the prior art does not provide a means to dynamically quantify compensation inaccuracies, resulting from the particular drift compensation technique used. These inaccuracies, inherent in the above compensation techniques, may vary widely, depending on specific operating conditions, such as: ambient temperature; number and location of temperature compensation/sensing devices in the system; thermal and time drift homogeneity among system components; time elapsed since a reference calibration (where relevant); and system warm-up status. In the above prior art, a general specification based upon a combination of individual system component drift tolerances, taken as a whole, can conceivably be computed to account for temperature and time drift limitations of a system. Based upon a given calibration, and/or compensation technique, over an intended temperature span, such a computation could be used to provide a limitation to achievable resolution in the above prior art, in a given operating environment, for a given calibration/compensation technique, over an expected temperature range of operation. However, in order to be reliable, such a computation would need to take into consideration factors which include long term time-drift characteristics of active and passive components, including amplifiers, and mixed-mode devices, such as analog/digital converters, as well as time drift of passive components, such as resistors and thermistors. An additional source of error to take into consideration, in devices expected to perform to specification when they are turned on, includes time versus drift behavior during system warm-up. Consequently, although a general specification for the ability of a technique to compensate for component time and temperature drift may be calculated, by combining manufacturer supplied drift specifications for relevant components, the actual value of errors associated with uncompensated component drift, in a given compensation technique, may change significantly, depending on the above factors, so that rather than being optimized, based upon current operating conditions, such a calculated specification must be set high enough to anticipate worst case conditions.
Such calculated resolution limits, for a given system, over a given temperature span, are often used to estimate performance in bridge measurement systems. However, it would be highly advantageous in detecting extremely small variations, approaching the limitations of modern electronics, to dynamically use limitation information that improves upon such absolute estimates, whenever possible. For example, system limitation specifications can be significantly enhanced by using such information as: elapsed time since a last reference calibration; empirically determined tolerance of temperature versus offset drift curves, over time; and elapsed time since power-up. In the above prior art, no means is provided to dynamically and efficiently account for collective circuit limitations, associated with a drift compensation method, in such a way as to provide an accurate, instantaneous indication, or continuous system control, reflecting optimum achievable system accuracy, under the drift compensation method, and based upon current operating conditions.
In spite of advances in bridge measurement systems, and in particular, high resolution temperature difference measurement systems, there remains a need for a high resolution differential thermometer, utilizing a minimum of low cost components, consuming minimal power (permitting battery powered operation), and operable from a single-ended power supply, that provides accurate temperature compensation, and that can be calibrated during normal operation, for temperature drift, and time drift of system components, without significantly interrupting operation in the field, and that sufficiently compensates for time drift of passive components, such as thermistors, so that a specified system resolution, on the order of micro-degrees centigrade, is achievable. Additionally, there remains a need for such a system which provides an instantaneous indication of system resolution limitations, based upon current operating conditions, which can be reported to the user, or employed to continuously and automatically effect system reporting in such a way as to dynamically provide optimal resolution, rather than using a single specification based upon a combined estimate of expected system tolerances.